Fragmentation Explained in 20 Minutes 

From my understanding, they are talking about an index in a database, so we want to have data stored in an index in some order (like names sorted alphabetically). But the index is stored under the hood in these 8kb pages, and inside the pages SQL Server does not implement a sorted order, it just has a list that holds the sort of the underlying data (idk if other database engines do it like that). So when people query the database, the query optimizer does not want to visit each page and see if record Simon is there (defeats the entire purpose of an index), it wants to immediately find the page with Simon, which means we need to do page splits from time to time to keep data sorted amongst different 8kb pages.